Practice parameter: Diagnosis of dementia (An evidence-based review)

PRACTICE PARAMETER: DIAGNOSIS OF DEMENTIA (AN EVIDENCE-BASED REVIEW) Report of the Quality Standards Subcommittee of the American Academy of Neurology D.S. Knopman, MD; S.T. DeKosky, MD; J.L. Cummings, MD; H. Chui, MD; J. Corey–Bloom, MD, PhD; N. Relkin, MD, PhD; G.W. Small, MD; B. Miller, MD; and J.C. Stevens, MD Article abstract—Objective: To update the 1994 practice parameter for the diagnosis of dementia in the elderly. Background: The AAN previously published a practice parameter on dementia in 1994. New research and clinical developments warrant an update of some aspects of diagnosis. Methods: Studies published in English from 1985 through 1999 were identified that addressed four questions: 1) Are the current criteria for the diagnosis of dementia reliable? 2) Are the current diagnostic criteria able to establish a diagnosis for the prevalent dementias in the elderly? 3) Do laboratory tests improve the accuracy of the clinical diagnosis of dementing illness? 4) What comorbidities should be evaluated in elderly patients undergoing an initial assessment for dementia? Recommendations: Based on evidence in the literature, the following recommendations are made. 1) The DSM-III-R definition of dementia is reliable and should be used (Guideline). 2) The National Institute of Neurologic, Communicative Disorders and Stroke–AD and Related Disorders Association (NINCDS-ADRDA) or the Diagnostic and Statistical Manual, 3rd edition, revised (DSM-IIIR) diagnostic criteria for AD and clinical criteria for Creutzfeldt–Jakob disease (CJD) have sufficient reliability and validity and should be used (Guideline). Diagnostic criteria for vascular dementia, dementia with Lewy bodies, and frontotemporal dementia may be of use in clinical practice (Option) but have imperfect reliability and validity. 3) Structural neuroimaging with either a noncontrast CT or MR scan in the initial evaluation of patients with dementia is appropriate. Because of insufficient data on validity, no other imaging procedure is recommended (Guideline). There are currently no genetic markers recommended for routine diagnostic purposes (Guideline). The CSF 14-3-3 protein is useful for confirming or rejecting the diagnosis of CJD (Guideline). 4) Screening for depression, B12 deficiency, and hypothyroidism should be performed (Guideline). Screening for syphilis in patients with dementia is not justified unless clinical suspicion for neurosyphilis is present (Guideline). Conclusions: Diagnostic criteria for dementia have improved since the 1994 practice parameter. Further research is needed to improve clinical definitions of dementia and its subtypes, as well as to determine the utility of various instruments of neuroimaging, biomarkers, and genetic testing in increasing diagnostic accuracy. NEUROLOGY 2001;56:1143-1153 Introduction. Mission statement. The Quality Standards Subcommittee of the American Academy of Neurology (AAN) is charged with developing practice parameters for physicians. This evidence-based review addresses major issues in the diagnosis of dementia. Background and justification. Dementia is a common disorder in the elderly, involving as many as 10% of those over 65 years of age. The AAN previously published a practice parameter on dementia in 1994,1 and since that time many new clinical and research developments have occurred. The purpose of the current practice parameter is to highlight and to update major areas of current interest and investigation in the diagnosis of dementia in the elderly. It is not intended to serve as a comprehensive review of the differential diagnosis of dementia.

The appointment of authors for this guideline was done in cooperation with the Alzheimer’s Association and overlaps significantly with the membership of the Medical and Scientific Advisory Council and the Board of Directors of the association. The Alzheimer’s Association agrees with the content of this paper in all important regards. This guideline has been endorsed by the American Association of Neuroscience Nurses and the American Geriatrics Society. From the Department of Neurology (Dr. Knopman), Mayo Clinic, Rochester, MN; Departments of Neurology and Psychiatry (Dr. DeKosky), University of Pittsburgh, PA; Departments of Neurology and Psychiatry & Biobehavioral Science (Dr. Cummings), University of California at Los Angeles; Department of Neurology (Dr. Chui), University of Southern California, Los Angeles; Department of Neurosciences (Dr. Corey–Bloom), University of California at San Diego; Department of Neurology and Neuroscience (Dr. Relkin), New York Presbyterian–Weill Cornell Medical College; Neuropsychiatric Institute and Hospital (Dr. Small), University of California at Los Angeles; Department of Neurology (Dr. Miller), University of California at San Francisco; Lutheran Medical Office (Dr. Stevens), Fort Wayne, IN. Approved by the AAN Quality Standards Subcommittee, November 11, 2000. Approved by the AAN Practice Committee January 6, 2001. Approved by the AAN Board of Directors February 24, 2001. Received July 6, 2000. Accepted in final form February 2, 2001. Copyright © 2001 by AAN Enterprises, Inc.

Diagnosis of Dementia

Page 2

Clinical question statement. The Diagnosis of Dementia Committee reviewed the issues, problems, and challenges related to the diagnosis of dementia. Based on work that has been published since 1994, the Committee formulated four questions to be addressed in the practice parameter: 1. Are the current criteria for the diagnosis of dementia reliable? 2. Are current diagnostic criteria sufficiently accurate to establish a diagnosis for the prevalent dementias in the elderly? 3. Do laboratory tests improve the accuracy of the clinical diagnosis of dementing illness? 4. What comorbidities should be evaluated in elderly patients undergoing an initial assessment for dementia? Process. Panel selection. A group of clinicians from various disciplines with extensive experience in diagnosing and caring for patients with dementia was assembled. The group was charged with focusing on the diagnosis of dementia. Committee members disclosed any real or potential conflicts of interest. Other work groups formulated practice parameters on the detection of dementia1 and the management of dementia.2 Literature review process. A literature search was conducted using MEDLINE, Excerpta Medica, and BIOSIS. The search included articles published from January 1985 through November 1999. The search strategy sought only studies published in English and studies on human disease. The principal search term was dementia. Other terms entered into the search included neuroimaging, diagnostic techniques, diagnostic imaging, biologic markers, CSF, diagnostic errors, differential diagnosis, and neuropsychologic tests. The original search yielded 1,175 articles of which approximately 300 articles were identified as relevant to our search questions. For articles on AD, we included only those based on more than 25 patients. For the less common dementias there was no minimal sample size. An additional 300 articles not identified by the literature search strategy, including ones published after the initial search was conducted, were submitted by committee members or obtained from bibliographies of articles identified in the search. Each article was classified based on the quality of evidence (Class I through IV, table 1). After review of the evidence, recommendations were drafted, reviewed by all committee members, and identified as a Practice Standard, Guideline, or Option (table 2). When appropriate, data on specificity, sensitivity, and other numerical measures of diagnostic precision were extracted from the articles and placed in tables. Final inclusion of articles in this Practice Parameter was based on consensus of the Committee that they were relevant and informative in consideration of the four questions. In addition to the review and final approval by the Quality Standards Subcommittee and the Practice Committee of the AAN, this Practice Parameter was reviewed by AAN members who had identified themselves as interested reviewers, by the Geriatric and Behavioral Neurology Sections of the AAN, by representatives of the American Geriatrics Society, and by representatives of the Alzheimer’s Association. Analysis of the evidence. Are the current criteria for the diagnosis of dementia reliable? The diagnostic formulations of dementia that are the most widely used in North America are based on definitions contained in the National Institute of Neurologic, Communicative Disorders and Stroke–AD and Related Disorders Association (NINCDS-ADRDA) Work Group,2 the Diagnostic and Statistical Manual, 3rd edition, revised (DSM-IIIR),3 and the Diagnostic and Statistical Manual, 4th edition (DSM-IV).4 The DSM-IIIR states: The essential feature of Dementia is impairment in short- and long-term memory, associated with impairment in abstract thinking, impaired judgment, other disturbances of higher cortical function, or personality change. The disturbance is severe enough to interfere significantly with work or usual social activities or relationships with others. The diagnosis of Dementia is not made if these symptoms occur. . .in Delirium. . .3

The DSM-IIIR definition of dementia has good to very good reliability (kappa’s ranging from 0.5 to 0.9).5-7 The closely related NINCDS-ADRDA and DSM-IV definitions of dementia have not been subjected to assessment of reliability. Practice recommendations. The DSM-IIIR definition of dementia, which is identical to the DSM-IV definition, is reliable and should be used routinely (Guideline). Are current diagnostic criteria able to establish a diagnosis for the prevalent dementias? Alzheimer’s disease. The reliability of the diagnosis of AD is moderate (generalized kappa’s in the 0.51 to 0.73 range) in Class II studies.8-10 When standardized clinical diagnostic criteria are used, interrater reliability6 and consistency of diagnosis between the initial visit and 1-year follow-up is high (95% in Forette11). There are 13 studies, 3 Class I12-14 and 10 Class II,9,15-22 that have addressed the diagnostic accuracy of the clinical diagnosis of AD using neuropathologic confirmation as the "gold standard." Both the DSM-IIIR "Dementia of the Alzheimer type" (DAT) and the NINCDS-ADRDA "probable" AD definitions achieved either good sensitivity (average across cited studies = 81%, range 49 to 100%) for AD at the expense of specificity (average across cited studies = 70%, range 47 to 100%) or vice versa in the majority of the cited studies. A diagnosis of "possible" AD achieved very high

Diagnosis of Dementia

Page 3

Table 1 Classification of evidence Class I II

III IV

Description Evidence provided by a well designed prospective study in a broad spectrum of persons with the suspected condition, using a “gold standard” for case definition, in which test is applied in a blinded evaluation, and enabling the assessment of appropriate tests of diagnostic accuracy. Evidence provided by a well designed prospective study of a narrow spectrum of persons with the suspected condition, or a well designed retrospective study of a broad spectrum of persons with an established condition (by “gold standard”) compared with a broad spectrum of controls, in which test is applied in blinded evaluation, and enabling the assessment of appropriate tests of diagnostic accuracy. Evidence provided by a retrospective study in which either persons with the established condition or controls are of a narrow spectrum, and in which test is applied in a blinded evaluation. Any design in which test is not applied in blinded evaluation OR evidence provided by expert opinion alone or in descriptive case series (without controls).

Table 2 Definitions for practice recommendations based on classification of evidence Recommendation Standard Guideline Practice Option Practice Advisory

Description Principle for patient management that reflects a high degree of clinical certainty (usually this requires Class I evidence that directly addresses the clinical question, or overwhelming Class II evidence when circumstances preclude randomized clinical trials). Recommendation for patient management that reflects moderate clinical certainty (usually this requires Class II evidence or a strong consensus of Class III evidence). Strategy for patient management for which the clinical utility is uncertain (inconclusive or conflicting evidence or opinion). Practice recommendation for emerging and/or newly approved therapies or technologies based on evidence from at least one Class I study. The evidence may demonstrate only a modest statistical effect or limited (partial) clinical response, or significant cost-benefit questions may exist. Substantial (or potential) disagreement among practitioners or between payers and practitioners may exist.

sensitivity (average across 4 studies = 93%, range 85 to 96%) but at the price of specificity (average across 4 studies = 48%, range 32 to 61%),13,14,17,22 reflecting the many features that non-AD dementias share with AD. Vascular dementia (VAD). Four criteria for vascular dementia that are currently used include the State of California AD Diagnostic and Treatment Centers criteria (the "California" criteria),23 the National Institute of Neurologic Disorders and Stroke and the Association Internationale pour la Recherche et l’Enseignement en Neurosciences (NINDS-AIREN) criteria,24 the Hachinski Ischemic Score (HIS) as modified by Rosen,25,26 and those found in DSM-IV.4 In studies that compared clinical diagnoses and neuropathologic findings, the NINDS-AIREN and the California criteria (as well as DSM-IIIR) had very low sensitivity but higher specificity. Only one Class I study12 reported the sensitivity (43%) and specificity (95%) of a published criteria, NINDS-AIREN, for VAD. Four Class II studies with patient samples drawn from referral cohorts15-17,27,28 reported sensitivity and specificity of the diagnosis of vascular dementia with any criteria. With one exception, their results had the same diagnostic accuracy as the population-based studies, with low sensitivity (average across 5 studies = 50%, range 20 to 89%) but higher specificity (average across 5 studies = 87%, range 64 to 98%) for the HIS, DSM-IIIR, NINDS-AIREN, or California clinical criteria. A retrospective (Class II) study27 in which the HIS showed better sensitivity and specificity (both 89%) was the one analysis of six in which the diagnosis of vascular dementia appeared both sensitive and specific. Recent neuropathologic analyses12,14 offer a perspective on the difficulty in correctly diagnosing cerebrovascular pathology in dementia. Rather than considering vascular dementia as simply present or absent, these studies distinguished between "some or any" vascular lesions versus "pure" vascular pathology, i.e., the circumstance in which vascular pathology was both sufficient to account for cognitive symptoms and unaccompanied by other pathology. Some vascular pathology exists in 29 to 41% of dementia cases coming to autopsy in population-based cohorts, even though pure vascular pathology accounted for dementia in only 9 to 10%.12,14 The Hachinski Ischemic Score, while lacking neuroimaging criteria, may be more suitable for identifying the majority of dementia patients with vascular dementia, i.e., those with at least some cerebrovascular pathology,27 because of the low sensitivity of the NINDS-AIREN and California criteria.12,15-17,28 Dementia with Lewy bodies (DLB). DLB has been defined clinically by the presence of dementia, gait/balance disorder, prominent hallucinations and delusions, sensitivity to traditional antipsychotics, and fluctuations in alertness.29 Two Class II studies evaluated the interrater reliability of the diagnosis of DLB based on the Consortium for DLB diagnostic criteria29 and found it to be relatively low.10,30 One Class I study12 investigated the diagnostic accuracy of DLB criteria against neuropathologic findings and found that sensitivity was low (22%) but specificity (100%) was high. Five Class II studies also showed low sensitivities (average across 5 studies = 58%, range 34 to 75%) but higher specificities (average across 5 studies = 87%, range 71 to 94%) for the diagnostic criteria of Consortium for DLB.10,30-33 In a prospective clinical study based on a DLB case registry with neuropathologically confirmed cases, hallucinations, depression, delusions, and delusional misidentification were all significantly higher in patients with DLB than AD.34 However, all of these features occurred in patients with AD

Diagnosis of Dementia

Page 4

as well. The presence of visual hallucinations and delusional misidentification as early symptoms showed sensitivities and specificities of >50% but 20 and 81.2% for MMSE score 20 and 100% for MMSE score